Skip to content

GRBM weights are initialized arbitrarily and aren't robust #48

@kevinchern

Description

@kevinchern

@VolodyaCO brought this up at https://github.com/dwavesystems/dwave-pytorch-plugin/pull/47/files#r2577897489 .
Empirically, the current weight initialization scheme tends to result in optimizer getting trapped in poor local optima.
We should run a simple experiment to pick a better initialization scheme.
For what it's worth, I've found initializing weights to 0 to be quite robust.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requestinvestigation-neededExpecting a more thorough investigation to resolve this issue

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions